Stats

  • Locked Locked
  • Replies 3
  • Subscribers 124
  • Views 13958
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Why does integrated tdnoise grow without bounds?

itos

Hello,

I have an integrator circuit that is reset periodically. Right before the reset, a sample is taken on which I would like to calculate the noise.

Conceptually, after the reset, the system is not in steady state and the output (noise) variance will grow until the system reaches steady-state.

The noise analysis gives me the steady state result. The output noise is attached at the end and the total integrated noise is 125uVrms.

Now I run pss+pnoise with tdnoise and plot the total integrated noise vs. time. This is what the outcome looks like:

Except for the y-axis scaling exactly as expected! The waveform of the reset signal is clearly visible: The system is reset between about 0 and 70ns, then the integrator starts and the total noise increases. It reaches a maximum at 250ns=0ns; then it is reset again.

However, this shows the plot for different values of "Integration Stop Frequency". 4MHz, 1PHz, 10PHz, 100PHz, 1000PHz:

Dependening on which value I choose for the stop frequency I can make the total noise shift arbitrarily. This does not make sense to me. For noise it's clearly visible that the system is bandlimited so the total noise must converge to a fixed number. Furthermore, it must be always smaller than the value in steady state (125uVrms from .noise). Particularly, when I make the number of sidebands (and maxacfreq in pss) very high, the number must converge to the one from .noise.

Why is this? Are my settings wrong?

I use pss; beat freq=4MHz, number harmonics=49, moderate, maxacfreq=1G

pnoise: Start: 1, Stop: 100G (some very high number), 10pts per decade. Maximum sidebands: 51.

Finally I use ADE L -> Direct Plot -> tdnoise -> Integ Output Noise, Total Noise, Start Frequency 1 Hz and Stop Frequency I vary from 1 MHz .... 10000PHz with difference results (as above).

As can be inferred from the plots, the control signals run at 4 MHz with ~21:6 duty cycle.

PS: This is the .noise output. Total integrated: 125uVrms.

  • Andrew Beckett

    When you use the "time domain" noise feature of pnoise (or the "PM" jitter mode; both are now called "jitter (sampled)" in the latest versions), the simulator inserts an ideal sampler at the output of the circuit which samples the circuit at a specific instant in time during the period (it may sample at multiple times, each is a "time event"). Anyway, this ideal sampler is operating at the PSS fundamental frequency, so that means that if you sweep beyond half the PSS fundamental in the pnoise analysis, you are causing the sampler to alias the noise and so it will double count, triple count etc the noise.

    Because of the ideal sampler, you are capturing all the noise because it folds the noise itself into the band up to half the PSS fundamental.

    So do not sweep beyond the PSS fundamental frequency divided by two. Similarly do not integrate the noise beyond PSSfund/2 for the same reason.

    Regards,

    Andrew.

  • itos
    in reply to Andrew Beckett

    Great, thank you, that makes sense! I used now VAR("fint")/2 for pnoise end range as well as for the total integrated noise integral limit.

    However, I have still trouble to reproduce the 125uV from .noise with .pnoise. I should get these in the limit when I wait until the system has sufficiently settled. In the plots above I do not wait too long enough so I set the clock fint to 476kHz such that the system settles.

    Based on the .noise output I would expect that a maxacfreq of 10-100 MHz should bring me into the range of 125uV. However, with:

    • maxacfreq=10M, numberOfHarmonics=10MHz / (476kHz/2) = 42 I get 57uVrms
    • maxacfreq=100M, numberOfHarmonics=420: 85uVrms
    • maxacfreq=1GHz, numberOfHarmonics=4200: ... the simulation runs forever, I think the number of harmonics is prohibitive

    Based on the .noise output, what would be the proper value for maxacfreq and numberOfHarmonics?

  • Andrew Beckett
    in reply to itos

    Are you using the fullspectrum option on pnoise? Then you should only need to specify a small number of harmonics in the pnoise analysis (to cover the flicker noise harmonics). In general you should not need to specify a large maxacfreq.

    If  you're not getting results that make sense, the best thing is to contact customer support so that we can take a look at the setup and the circuit itself to understand what's going on. It's going to be very hard to guess in the forums just from the small snapshot of the information you can provide here...

    Regards,

    Andrew

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.